About us
The Anthropomorphic Interfaces Laboratory develops companion robots and conducts research in the field of human-machine interfaces. We are developing the F-2 research robot, which can be used to study methods of communication between robots and humans. The robot consists of several software components: it accepts natural language text as input, constructs the meaning of this text, selects an emotional or rational response, and performs gestures characteristic of the selected response. It is possible to develop specified behavioural sequences for the robot, create behavioural templates, and use them in interactive experiments.
We are developing a software architecture for the robot that will allow it to appear human in social situations. The robot responds to statements and can read text files: novels or daily news. For each sentence, it constructs a syntactic tree and a semantic representation. It can store the semantics of texts in a database and search it for answers to questions. The robot compares each meaning with the premises of ‘scenarios’ that allow it to respond rationally or emotionally to incoming statements. With the help of separate software components, the robot can react to the user’s facial expressions or gaze, assemble a tangram puzzle with the user or play a computer game, react to the user’s touch, help the user learn foreign language words, etc.
In each of these interaction scenarios, the developer can create different communication strategies for the robot and test them in an experiment. The communication strategies selected in the experiments allow the robot to appear attractive and friendly to humans.
As part of a series of experiments, we investigated how a robot might behave when helping a person assemble a puzzle or teaching a person a foreign language. The robot has already taught students and schoolchildren Latin, English and Chinese. In a series of experiments, we studied how humans perceive the robot’s gaze, smiles, and other actions. In several experiments, we also assessed whether a robot should appear strict or emotional when teaching humans, and which behavioural strategies are better perceived by humans.
The laboratory invites students to complete coursework and dissertations.
Research areas
1.Development of a semantic parser for automatic text analysis
- Research into verb control models to improve parser efficiency;
- Comparison of semantic frames in texts from different sources;
- Typology of verb control models for text synthesis based on semantic representation;
- Study of semantic fields and refinement of semantic markup;
- Analysis and experimental study of humour mechanisms in robot-human communication;
- Mechanisms of linguistic homonymy for automatic creation of humorous texts;
- Modelling of metaphor mechanisms in automatic text analysis.
2.Development of a natural language inference system for a companion robot
- Research into the means of constructing implicit meanings in verbal communication.
3.Development of nonverbal behaviour patterns for a companion robot
- Assessment of the impact of smiles/nods/hesitation markers/internal speech markers of a robot companion on the listener;
- Study of non-verbal means of establishing and maintaining joint attention in robot-human communication;
- Assessment of trust in the robot depending on the non-verbal communication means used;
- The influence of message semantics on the choice of iconic gestures;
- Assessment of the direction of the temporal axis in human gestures.
